Click to edit Master title style - DTIC · based on: • derived model • starting conditions •...

1

Click to edit Master title style

U.S. Army Armament Research, Development and Engineering Center

Method and Process For The Creation of Modeling and Simulation Tools for Human Crowd BehaviorMr. GLADSTONE V. REID JR.Presented to the 82nd Military Operations Research Society SymposiumJune 4-6, 2014

Distribution Statement A - Approved for Public Release

Report Documentation Page Form ApprovedOMB No. 0704-0188

Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering andmaintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, ArlingtonVA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if itdoes not display a currently valid OMB control number.

1. REPORT DATE 23 JUL 2014

2. REPORT TYPE Conference Presentation

3. DATES COVERED 00-00-2011 to 00-00-2014

4. TITLE AND SUBTITLE Method and Process For The Creation of Modeling and Simulation Toolsfor Human Crowd Behavior Presented at the Virtual 82nd MilitaryOperations Research Society Symposium July 23-24 2014

5a. CONTRACT NUMBER

5b. GRANT NUMBER

5c. PROGRAM ELEMENT NUMBER

6. AUTHOR(S) Gordon Cooke; Gladstone Reid; Robert DeMarco; Elizabeth Mezzacappa

5d. PROJECT NUMBER

5e. TASK NUMBER

5f. WORK UNIT NUMBER

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) Army, ARDEC, Target Behavioral ResponseLaboratory,RDAR-EIQ-SD,Building 3518,Picatinny Arsenal,NJ,07806-5000

8. PERFORMING ORGANIZATIONREPORT NUMBER

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)

11. SPONSOR/MONITOR’S REPORT NUMBER(S)

12. DISTRIBUTION/AVAILABILITY STATEMENT Approved for public release; distribution unlimited

13. SUPPLEMENTARY NOTES

14. ABSTRACT Commanders need tools for planning, decision support, and analysis related to crowd management withnon-lethal weapons. This problem requires predictive crowd modeling and simulation tools for forecastinghuman behavior. The Target Behavioral Response Laboratory (TBRL) has worked to develop newmethods to provide modeling and simulation operational planning tools to provide commanders with thecapability to predict crowd response to military control force tactics, techniques, and procedures. TheTBRL has developed processes based on behavioral science experimental methods and objectivemeasurements under controlled conditions. These methods and measures include motion capture ofsubjects in a laboratory environment to derive coefficients for equations which quantitatively model acrowd’s behavior, which are can be validated at more than one point in the process. This technique canmodel a crowd’s path deviation response to different weapon, device, or control force emplacements as thecrowd moves towards an objective of positive valence, which might aid a commander in choosing andpositioning those weapons. The TBRL has gathered a large data set from a variety of human crowdexperiments using non-lethal energies and devices to determine effectiveness. These data serve as the inputand validation tools for model building. The process comprises of several modules that work together toproduce simulated data for crowd locomotive behavior; estimating crowd responses to several non-lethaltechnologies and their surrogates. The process yields models that accurately estimate crowd behavior forbaseline and a Medium Range Acoustic Device condition and partially estimates crowd behavioralresponse for area denial technology (ADT) and hand-held standoff non-lethal weapons. Building on humanbehavioral data allows the model to capture the behaviors, range, variability, probability, and uncertainty,the dynamic nature of human behaviors. Starting with the gathering of data on human behavior allows oneto configure the test to gather data on the information relevant to building the models and simulation toanswer the commander???s question. This process increases the availability of data on human behavior,facilitates the development of architecture-free, reusable, composeable, interoperable human behavioralmodels and simulation, and provides objective verification and validation metrics. It provides decisionsupport to commanders, specifically with quantitative information based on human behavioral datagathered under controlled conditions.

15. SUBJECT TERMS Human Behavior Modeling and Simulation, Mathematical Model, Computational Model, Non-lethalWeapons, Human Experimentation, Empirical Data, Behavioral Models, Target Behavioral ResponseLaboratory, Crowds, Predictive Crowd Modeling, Motion Capture

16. SECURITY CLASSIFICATION OF: 17. LIMITATIONOF ABSTRACT

Public Release

18. NUMBEROF PAGES

26

19a. NAME OFRESPONSIBLE PERSON

a. REPORT unclassified

b. ABSTRACT unclassified

c. THIS PAGE unclassified

Standard Form 298 (Rev. 8-98) Prescribed by ANSI Std Z39-18

2


2Unclassified

• Introduction– ARDEC– Crowd Behavior Research at TBRL

• Data Collection– Experimental Conditions

• Modeling & Simulation Process– Model Building– Simulation– Crowd Metrics Analysis– Model Validation

• Results– Model Verification– Model Validation

• Summary • Conclusions

Overview

3


3Unclassified

Research, Development and Engineering Command, RDECOMMr. Dale Ormond

Headquarters, Department of the Army

Army Materiel Command, AMCGen. Dennis L. Via

Armament Research, Development and Engineering Center, ARDECDr. Gerardo J. Melendez

PEO AmmunitionBG John J. McGuiness

Joint Munitions & Lethality LCMCBG Kevin O’Connell

• Program Executive Office Combat Support and Combat Service Support• Program Executive Office Ground Combat Systems• Program Executive Office Soldier

TACOM LCMCMG Michael J. Terry

Assigned/Direct SupportCoordination

US Army - ARDECUS Army - ARDEC

Assistant Secretary of the ArmyAcquisition, Logistics and Technology

Ms. Heidi Shyu

4


4Unclassified

Advanced Weapons:Line of sight/beyond line of sight fire; non line of sight fire; scalable effects; non-lethal; directed energy; autonomous weapons

Ammunition:Small, medium, large caliber; propellants; explosives; pyrotechnics; warheads; insensitive munitions; logistics; packaging; fuzes; environmental technologies and explosive ordnance disposal

Fire Control:Battlefield digitization; embedded system software; aero ballistics and telemetry

ARDEC provides the technology for over 90% of the Army’s lethality and a significant amount of support for other services’ lethality

ARDEC’s Role

RESEARCH DEVELOPMENT PRODUCTION FIELD SUPPORT DEMILITARIZATION

5


5Unclassified

Introduction

Military Need for Crowd Behavior Research

– The motivations underlying adversarial behavior

– Behavior of contested populations

– How the behavior of populations varies cross-culturally

– What is innate human behavior that extends across cultural boundaries?

6


6Unclassified

• Human behavior can be explained as attractions and repulsions toward and away from goals (Lewin, 1935)

• Crowd Behavioral Test Bed used to gather:– locomotive – psychosocial – effectiveness data

• Data gathered to develop models that use vector regression methods to identify attributes of a crowd that influence predictive variables

Introduction

Crowd Behavior Research at TBRL

7


7Unclassified

• Crowds with up to 25 subjects during a trial throwing simulated rocks into a linear target while the target was defended by a non-lethal device:

– No Defense (Baseline)

– MRAD

– Handheld stand-off NLW operated by Control Force

• Simulated Projectile Weapon

• Simulated Handheld Directed Energy NLW (VDE)

– Simulated Invisibly located Directed Energy NLW (IDE)

Data Collection

Experimental Conditions

8


8Unclassified

• MATLAB modules for developing software to:

– Automate the creation of mathematical models for motion of individuals in a crowd from the data collected in TBRL crowd experimentation

– Run simple simulation of the above model to generate simulated data

– Calculate crowd level metrics from laboratory and simulated data

– Compare simulated data with observed data

M & S Process

Model Building

9


9Unclassified

• Pre-Process Module

– Separates modeling and validation data

– Restructures the location data file• Data for each participant at sample time is grouped and

appended to preceding time steps

• Input Module

– Parses the formatted data from the previous step into three elements: • Header vector • Output matrix• Predictor matrix

M & S Process

Model Building

10


10Unclassified

• Modeling Module

– Accepts the predictor and output matrices

– Performs non-linear regressions to determine the relationship between predictor/input and predicted/output variables. • [beta, r, J, COVB, mse] = nlinfit(X, y, fun, beta0)

• Regression of velocity vectors in both ‘X’ and ‘Y’ axes against the predictors, generating model coefficients for change in location in ‘X’ and ‘Y’ coordinates

– Model errors were fit to Weibull distribution

M & S Process

Model Building

11


11Unclassified

• Simulation Module

– Execute a time stepped simulation of each subject’s location based on: • derived model • starting conditions • average time between samples • duration of simulation

– Calculate the following for each subject at each time step and appended to a simulated data file• delta distance traveled • delta position• velocities

– Incorporate Control Force effects by transforming the coordinates of the baseline model to fit that of the CF model

M & S Process

Simulation

12


12Unclassified

• Crowd Metric Module

– Calculates aggregate metrics of crowd behavior as a whole

• Leading edge (LE): location of the forward most crowd member

• Trailing edge (TE): location of the crowd member that is furthest back

• Centroid: location of the crowd member that is midway between the LE and TE

• Geometric center: mean of the LE & TE

• Dispersion: average displacement in the ‘X’ and ‘Y’ direction

M & S Process

Crowd Metric Analysis

13


13Unclassified

• Model Comparison

– Compares, statistically, the crowd metrics of simulated and observed data

• two sample Kolmogorov-Smirnov (K-S) goodness of fit (GOF) test

• Determine if the simulated data follows the same distribution as the observed data, the asymptotic p-value, and k-statistic

– 5% significance level

– Unequal alternative hypothesis test

M & S Process

Model Validation

14


14Unclassified

• RMSE comparison

– Comparison between the expected values calculated from the regression equation and the data observed in the laboratory

Results

Model Verification

Root Mean Square Error: RMSE

15


15Unclassified

• Graphical Comparison

Results

Model Verification

16


16Unclassified

• Graphical Comparison Cont’d

Results

Model Verification

17


17Unclassified

• K-S GOF Test

Results

Model Validation

18


18Unclassified

• K-S GOF Test - Baseline

Results

Model Validation

19


19Unclassified

• K-S GOF Test - MRAD

Results

Model Validation

20


20Unclassified

• K-S GOF Test - Projectile Weapon

Results

Model Validation

21


21Unclassified

• There are several identified areas for improvement which include:

– Using additional or new regression methods that yield more accurate models reflected by smaller RMSE

– Accounting for error during simulation and making necessary adjustments to produce data that more closely represent observed data

– The incorporation of crowd metrics for the observed data set into the model development process to improve the goodness of fit for the observed and simulated data

– Using additional goodness of fit tests to validate the model

Conclusion

22


22Unclassified

• The model development process created has successfully – created models from Laboratory collected data– simulated models over time to generate crowd data per condition– performed analysis of observed and simulated data– compared simulated and observed data against crowd behavioral

measures

• The process allows for quantitative means for validation of both the mathematical and computational models against empirical data

• The models created for the baseline and the MRAD weapon conditions successfully estimated crowd locomotive behavior

• Some of the models created still need refinement based on the graphical comparison of the simulated and observed data in addition to the K-S GOF test

Summary

23


23Unclassified

Target Behavioral Response Laboratory MORSS Presentations

• Virtual Employment Test Bed: Operational Research and Systems Analysis to Test Armaments Designs Early in the Life Cycle

• Method and Process for the Creation of modeling and Simulation Tools for Human Crowd Behavior

• Squad Modeling and Simulation for Analysis of Materiel and Personnel Solutions

• The Squad Performance Test Bed• Crowd Characteristics and Management with Non-Lethal Weapons: A

Soldier Survey• Effectiveness Testing and Evaluation of Non-lethal Weapons for Crowd

Management• Effects of Control Force Number, Threat, And Weapon Type on Crowd

Behavior

24


24Unclassified

Questions?

US Army - Target Behavioral Response Lab

Mr. Gladstone V. Reid Jr.Picatinny Arsenal, NJ

[email protected]

Questions & Answers

25


25Unclassified

BACKUP SLIDES

26


26Unclassified

Data Collection

Crowd Behavior Test Bed

Courtesy Vicon

Click to edit Master title style - DTIC · based on: • derived model • starting conditions •...

Documents

Transcript of Click to edit Master title style - DTIC · based on: • derived model • starting conditions •...